Physio-Control E-mail Systems

Graph of various parameter during IMAP mailbox access

Mark Fromm, from Seattle Sage Group topic on 1/09/97

Comparison of two types of mailbox access:

• Server is Sun Sparc running Solaris 2.5. Sun IMAP4 deamon
• Client is accessing a 4 megabyte, 1000 message mail file, first using Sun Mailtool (NFS mount), then using IMAP4 mail client.
• Chart is of mail server data (CPU, ethernet packets, Disk I/O, and Load)
• Sequence of events is
• NFS mailtool read
• NFS mailtool read, modify, and write
• IMAP4 open, first time (which develops mailbox index #s)
• IMAP4 open, second open, read only
• IMAP4 open, third open, modify and write
• IMAP4 open, server based search of entire mailfile by selected criteria



This chart has much interesting data. The left hand side represents mail access via NFS mount, while the right side represents access to the same mailfile from the same machine using IMAP4.


Note the uppermost trace of CPU utilization. The large spike on the first IMAP access indicates the indexing and number generation that IMAP needs to perform on a mailfile. This is a one time operation. Subsequent accesses are lightweight in system usage.


The second from the upper trace is the number of ethernet packets. In particular, note the small amounts of data sent over the network interface when using IMAP - this is particuarly important when considering access via dial up lines.


The second from the bottom trace is disk I/O - again using IMAP we are more efficient, especially when comparing to the NFS read, delete, and write back after exiting the mail program.


Finally we have load, while admittedly higher using IMAP is only slighty higher except during the initial IMAP mailbox access. In one instance I performed a server based search by keyword of the entire mailfile. This essentially trades (and preserves) network bandwidth for some additional server processing.

GO BACK